Is this using custom skeleton tracking from the Kinect v2 depth map or are you using the MS SDK? Any validation of the accuracy and reliability?